Red Supergiants and the Past of Cygnus OB2

A&A 586, A46 (2016) Astronomy DOI: 10.1051/0004-6361/201527517 & c ESO 2016 Astrophysics

Red supergiants and the past of Cygnus OB2

F. Comerón1,?, A. A. Djupvik2, N. Schneider3,4, and A. Pasquali5

1 European Southern Observatory, 3107 Alonso de Córdova, Vitacura, Santiago 19, Chile e-mail: [email protected] 2 Nordic Optical Telescope, Apdo 474, 38700 Santa Cruz de La Palma, Spain 3 I. Physik Institut University of Cologne, 50937 Cologne, Germany 4 OASU/LAB-UMR 5804, CNRS, Université Bordeaux 1, 33270 Floirac, France 5 Astronomisches Rechen-Institut, Zentrum für Astronomie der Universität Heidelberg, Mönchhofstr. 12–14, 69120 Heidelberg, Germany Received 6 October 2015 / Accepted 2 December 2015

ABSTRACT

Context. Red supergiants are the evolved descendants of massive stars with initial masses between 7 and 40 M . Their brightness makes them easily detectable in the near infrared, making them useful probes of star formation that occurred several tens of Myr ago. Aims. We investigate the past star formation history of Cygnus OB2, the nearest very massive OB association, using red supergiants as a probe. Our aim is to confirm the evidence, found by previous studies, that star formation in the Cygnus OB2 region started long before the latest burst that gave rise to the dense aggregate of early O-type stars that dominate the appearance of the association at present. Methods. Near-infrared star counts in the Cygnus region reveal moderate evidence for a peak in the areal density of bright, reddened stars approximately coincident with Cygnus OB2. A total of 11 sources are found within a circle of 1◦ radius centered on the association, of which 4 are non-supergiants based on existing observations. Near-infrared spectroscopy is presented of the remaining seven candidates, including four that have been already classified as M supergiants in the literature. Results. We confirm the presence of seven red supergiants in the region and argue that they are probably physically associated with Cygnus OB2. Their location is roughly coincident with that of the older population identified by previous studies, supporting the sce- nario in which the main star formation activity in the association has been shifting toward higher Galactic longitudes with time. Their luminosities are compared with the predictions of evolutionary tracks with and without rotation to estimate the mass of their progenitors and ages. In this way, we confirm that massive star formation was already taking place in the area of Cygnus OB2 over 20 Myr ago, and we estimate that the star formation rate in the latest 6 Myr represents a six-fold increase over the massive star formation rate at the time when the progenitors of the current red supergiants were formed. Conclusions. The Cygnus OB2 association has a history of star formation extending into the past for at least about 20 Myr, probably dovetailing with the general history of star formation in the region that gave rise to other associations like the neighboring Cygnus OB9. The sustained massive star formation history also argues for a long lifetime of the giant molecular complex from which Cygnus OB2 formed, whose remnants constitute the present-day Cygnus X complex. Key words. stars: late-type – open clusters and associations: individual: Cygnus OB2

1. Introduction Many of these recent studies (see e.g. Comerón & Pasquali 2012; Wright et al. 2015, and references therein) show that the The Cygnus OB2 association is one of the prime targets for stud- star formation history of Cygnus OB2 extends well into the past ies of the upper end of the stellar mass function and of the inter- and that the currently densest and youngest part of the associa- action of massive stars with their surrounding molecular envi- tion, where the vast majority of its O stars are found, is contigu- ronment (e.g. Knödlseder 2003; Schneider et al. 2006; Reipurth ous to, and partly overlapping with, an area where B giants and & Schneider 2008). The stellar population of Cygnus OB2 is supergiants abound, indicating that many stars in the association a showcase of nearly all the varieties of stages that very mas- have already evolved away from the main sequence and that the sive stars undergo in their evolution. Following early studies that main star-forming sites have been shifting with time. The study hinted at the true richness of this significantly reddened asso- of this older component is made difficult by the effects of stel- ciation (Massey & Thompson 1991; Knödlseder 2000), several lar evolution, and in particular by the fact that its most massive efforts have been made to produce an increasingly complete cen- components have already disappeared as supernovae. sus of the massive population of the association (Comerón et al. A thus far unexplored probe of the massive star formation 2002; Comerón & Pasquali 2012; Wright et al. 2015). Parallel past of Cygnus OB2 is composed of cool supergiants, the de- to this, studies have investigated aspects such as its spatial ex- scendants of stars with original masses between ∼7 and ∼40 M tent, star formation history, lower mass content, or possible rela- (Hirschi 2010) that reach very low (Teff < 4000 K) photospheric tionship with neighboring associations (Hanson 2003; Comerón temperatures at some stages of their post-main sequence evo- et al. 2008; Drew et al. 2008; Wright et al. 2010; Comerón & lution. If present, red supergiants should be the brightest mem- Pasquali 2012; Wright et al. 2014). bers of Cygnus OB2 at infrared wavelengths and their spectra should make them easily recognizable; indeed, the high in- ? Visiting astronomer at the Vatican Observatory. frared luminosities of these stars make them a useful tool for

Article published by EDP Sciences A46, page 1 of8 A&A 586, A46 (2016) the identification and study of distant, highly reddened clusters (e.g. Messineo et al. 2014; Davies et al. 2007; Clark et al. 2009; Mengel & Tacconi-Garman 2007). At the same time, their presence and characteristics would set relevant lower limits on the past massive star formation activity of the region. In this paper we report on seven red supergiants located within or near the boundaries of Cygnus OB2 and consistent with being located at the same distance as the association. Four of them have previously been recognized and classified as M supergiants, but apparently their possible relationship with Cygnus OB2 had not been taken into consideration. The remaining three are confirmed as red supergiants in this work. By using the most recent synthetic evolutionary tracks we discuss the initial properties and ages of these stars, and are able to extract some conclusions on the distant star formation past of the pre- cursor of the present-day Cygnus OB2 association.

2. Source selection Galactic M supergiants (luminosity classes I–II) cover a range of luminosities spanning nearly two orders of magnitude starting at log(L/L ) ' 4 while being confined to a rather narrow range of temperatures between 3500 K and 4000 K (Levesque et al. 2005; Fig. 1. Color–magnitude diagram of stars with K < 8 in the field see Davies et al. 2013, however). Using the synthetic magnitudes of 1◦ radius around l = 79◦8, b = +0◦8. The filled circles are the candi- derived in the Geneva evolutionary tracks for solar metallicity of date M supergiants selected for spectroscopic observations and, except Ekström et al. (2012), the faint end in luminosity corresponds to for the brightest of them, the photometry has been obtained by us as de- an absolute magnitude MK ' −7.8, or an unreddened apparent scribed in Sect. 3.1. Photometry for the remaining stars has been taken magnitude K0 ' 3.0 at the distance of 1.45 kpc (distance modu- from the 2MASS Point Source Catalog. lus DM = 10.8; Hanson 2003; Rygl et al. 2012), which we adopt in this study. The bulk of Cygnus OB2 members listed in the recent compilation of Wright et al. (2015) has visual extinctions in We note that this definition of the boundaries of Cygnus OB2 the range 4 < AV < 7, which using the Cygnus OB2 extinction differs from that adopted by Wright et al. (2015), which uses a law derived by Wright et al. (2015) implies 0.35 < AK < 0.60. circle of smaller radius centered on the Trapezium-like system We thus expect M supergiants in Cygnus OB2 to be very bright VI Cyg 8, very close to the density peak. Our search area thus at infrared wavelengths, with K < 3.6. includes the evolved population, mostly adjacent to the area pop- Several compilations of intrinsic colors of M supergiants can ulated by the newest, early O-type members of the association. be found in the literature, with published values covering a fairly Our selection criteria provided a total of 11 candidates, wide range for a given spectral subtype. Straižys (1987) gave which are plotted in a near-infrared color-magnitude diagram in (J − K)0 values ranging from 0.91 to 1.20 in the M0I-M4I range, Fig.1. Four of them could be discarded as M supergiant can- while Ducati et al. (2001) quoted significantly different values, didates on the basis of previous studies: they are the K5 giant ranging from 0.85 to 0.64 in the same range of spectral types. HD 196241 (Haggkvist & Oja 1970), the M8e Mira variable We regard the values of the compilation of Ducati et al. (2001) in KZ Cygni (Cameron & Nassau 1956), the B5Ia+ supergiant this spectral range with some caution because not only are those VI Cyg 12 (Humphreys 1978) and the Bep star MWC 349 values notoriously lower than those of M spectral types of other (Brugel & Wallerstein 1979). luminosity classes, but they also become bluer with increasing Of the remaining seven candidates, four have previ- spectral subtype and their (H − K)0 indices remain negative over ously been confirmed as supergiants through spectroscopy the whole M supergiant spectral range. For the present study we in the visible by other authors: RAFGL 2600 (M1:Iab; have preferred to use the compilation of Tokunaga (2000), which Verhoelst et al. 2009), IRAS 20341+4047 (M3I; Imanishi et al. lists colors ranging from (J−K)0 = 0.79 for a spectral type M0Ia 1996), RAFGL 2605 (M4I:; Grasdalen & Sneden 1979), and to (J − K)0 = 1.07 for M4Ib. IRAS 20315+4026 (M0I; Imanishi et al. 1996). No classifica- We have thus searched the 2MASS all-sky point source cata- tion is available in the literature for the remaining three objects in log to select a sample of candidate M supergiants in Cygnus OB2 our selection, which are IRAS 20290+4037, IRAS 20263+4030, defined by the conditions K < 4.0, (J − K) > 1.1, thus restricting and IRAS 20249+4046. the selection to bright and significantly red, or reddened, stars, Although individual distance estimates for these stars are not and providing a virtually complete sample of M supergiants at available, the association of red supergiants with Cygnus OB2 is the distance of Cygnus OB2. The brightness limit in K excludes indirectly supported by the fact that the areal density of objects the entire early-type population of Cygnus OB2, whose bright- selected by the criteria described above peaks at the position of est members are expected to peak at K0 > 5 (Martins & Plez Cygnus OB2 when a broad range of Galactic longitudes is con- 2006) except for possible stars with strong near-infrared excess. sidered. Applying these selection criteria to areas of 1◦ radius The search area has been defined as a circle of 1◦ radius cen- that are centered at Galactic longitudes ranging from l = 69◦8 to tered on Galactic coordinates l = 79◦8, b = +0◦8 in consistency l = 89◦8, always at a Galactic latitude b = 0◦8, we find an av- with the boundaries of Cygnus OB2 proposed in Comerón et al. erage of 6.6 objects per area, with a standard deviation σ = 2.7. (2008) and Comerón & Pasquali (2012), which includes, but is The fact that we find 11 objects within the area of the same radius offset from, the current stellar density peak of the association. centered on Cygnus OB2 thus provides some evidence (within

A46, page 2 of8 F. Comerón et al.: Red supergiants and the past of Cygnus OB2

Table 1. Basic data of the candidate red supergiants.

Star ID α(2000) δ(2000) JHKS Sp. type Name 1 20:33:23.92 +40:36:44.2 6.984 4.887 3.902 M0I1 IRAS 20315+4026 2 20:28:06.19 +40:40:06.1 5.675 4.420 3.518 – IRAS 20263+4030 3 20:33:01.06 +40:45:40.4 3.8592 2.0262 1.2962 M4I: RAFGL 2605 4 20:30:51.53 +40:48:08.6 4.884 3.814 3.342 – IRAS 20290+4037 5 20:31:28.70 +40:38:43.3 4.131 2.883 2.328 M1:Iab RAFGL 2600 6 20:26:43.03 +40:56:26.8 5.085 3.645 3.095 – IRAS 20249+4046 7 20:35:54.73 +40:57:40.1 5.869 4.198 3.288 M3Iab:3 IRAS 20341+4047

Notes. (1) We adopted the classification of Imanishi et al. (1996). Rawlings et al. (2000) suggested that it might be an early star, but this is conclusively ruled out by the K-band spectroscopy presented here. (2) Photometry obtained from 2MASS, uncertain by ∼0.3 mag in each band. The star is too bright for the setup used in our observations, even after strong defocusing. (3) We adopted the classification of Imanishi et al. (1996), which is consistent with the near-infrared spectrum presented here. We note, however, that there are highly discrepant spectral classifications of this object (Rawlings et al. 2000). the limitations imposed by small-number statistics) of an excess at near-infrared wavelengths. The photometry of our targets is of very bright, red sources coincident with the association. It presented in Table1. must be noted that the Galactic longitude range we considered to estimate the density of such sources outside Cygnus OB2 includes other OB associations in Cygnus (Uyaniker et al. 2001) 3.2. Spectroscopy and thus possibly a certain contribution by red supergiants that belong to those associations. On the other hand, the members All seven stars in our sample, including those for which spec- of our sample of candidate supergiants have infrared colors that tral classifications as supergiants were already available in the imply extinctions in their direction similar to, and in some cases literature, were observed in the near-infrared K-band using the even higher than, those typical of Cygnus OB2 members, which NOTCam array camera and spectrograph at the Nordic Optical argues against a possible contamination of our sample by fore- Telescope (NOT) at the Observatory of Roque de los Muchachos ground late-type red giants. We therefore consider here that that (La Palma, Canary Islands, Spain). The spectra were obtained the red supergiants we discussed are all at the same distance as on the nights of 1 June, 2 August, and 2 September 2015. The Cygnus OB2. A0V star HIP 99719, observed with the same setup as our targets just before and after them, was used as a calibrator to remove telluric features. 3. Observations The NOTCam spectrograph with its Hawaii-1 HgCdTe array was used with the wide-field camera (0.23400/pix), Echelle 3.1. Imaging grism #1, a 0.600 wide slit, and the MKO K-band filter used as an The brightness of our targets means that some of them appeared order sorter. With this setup the dispersion is 4.1 Å/pix, giving a saturated in the 2MASS survey and that their cataloged 2MASS spectral resolution λ/∆λ ∼ 2100 at 2.2 µm. magnitudes were derived from the extrapolation of the unsatu- The data were obtained in an A-B-B-A dithering pattern rated wings of their point-spread function, whereby the uncer- along the slit. The very bright targets needed exposure times in tainties in the flux measurements reached 20%–30%. We there- the range 3.6 to 12 s per dither position, while for the fainter fore decided to reobserve all the stars in the J, H and KS-bands telluric standard calibrator we used 60 s, all obtained with the using the CAIN-3 near-infrared camera at the 1.5 m Carlos sample-up-the-ramp readout mode. When acquiring the stars on Sánchez telescope at the Observatorio del Teide (Tenerife, the slit, we added a narrow-band filter in the beam to avoid heavy Canary Islands, Spain). The observations were carried out in saturation of the detector, which otherwise could produce persis- service mode on the nights of 19 May and 2 June 2015. To tency effects in the spectra. Halogen lamp flats were obtained in avoid saturation the telescope was defocused before each obser- situ to better correct for fringing. vation and counts over the resulting stellar image were checked One-dimensional spectra of our targets and telluric calibra- to fall within the linear regime of the detector. Ten stacks of im- tion star were extracted using dedicated scripts based on IRAF ages were obtained for each star, each consisting of eight in- tasks and wavelength calibrated using the sky emission lines as dividual exposures of 0.25 s integration time, with small (1500) a built-in calibration source (Oliva & Origlia 1992). The tar- telescope offsets in between. The images were flat-fielded and get spectra were ratioed by that of the telluric calibration star sky-subtracted using as background sky the image obtained by and multiplied by an artificially generated blackbody spectrum combining the ten stacks uncorrected for the telescope dithers, at T = 10 000 K, approximately representing the spectrum of the using sigma clipping to remove the stellar images from the in- A0V telluric calibrator, to obtain a relative flux calibration. dividual frames. Four bright stars with well-determined magnitudes in 2MASS near our targets were observed just before and after to provide accurate flux calibration. Aperture photometry 4. Results was then performed on the reduced images of all the sources us- 4.1. Spectra classification ing an oversized aperture to include the entire flux from the defocused images. Contamination in such large apertures proved to Figure2 displays the spectra of all our M supergiant candidates, be negligible, as both our targets and the stars selected for flux including those for which a classification is already available in calibration are by far the brightest objects in their fields of view the literature.

A46, page 3 of8 A&A 586, A46 (2016)

(EW) of those features that were obtained using passbands centered on them. The reference continuum was taken on both sides of the feature and interpolated, with the exception of the CO(λ = 2.293 µm) molecular band for which two reference 1 continuum intervals on the shorter wavelength side were used for extrapolation, because there is no suitable continuum near the red edge of the feature. The positions and widths of the passbands used to measure the equivalent widths of the features 2 and their chosen continuum reference points are the same as described in Comerón et al. (2004). Uncertainties in the equivalent width measured in this way were estimated by repeating 3 the measurements with different choices of the continuum level. We thus estimate errors ∆EW(CO) ' 1.1 Å, and ∆EW(NaI) ' ∆EW(CaI) ' 0.8 Å. 4 Equivalent widths of the CO features of our candidates are consistently in the range of the galactic supergiants presented by Lançon & Wood (2000). This same range is also occupied by 5 oxygen-rich long-period variables, but the latter usually display deep and very broad water vapor absorption bands whose wings 6 dominate the appearance of the K-band spectrum at wavelengths shorter than ∼2.1 µm. These are clearly absent from our spectra. On the other hand, all the measurements of EW(CO) of red gi- 7 ants in the sample considered by Davies et al. (2007) fall below those that we obtain in our sample, even including red giants of types M7 or later, thus giving us further confidence that all the stars in our sample are true supergiants. We used our spectra to estimate spectral types according to their correlation with EW(CO) for supergiants presented by Davies et al. (2007). The blue edge of the passband we used to define EW(CO) is at a slightly shorter wavelength (2.289 µm) Fig. 2. K-band spectra of the candidate Cygnus OB2 M supergiants obtained with NOTCam at the Nordic Optical Telescope. The appearance than the band used by these authors. This is nevertheless nec- of the spectra is dominated by the CO bands longward from 2.293 µm, essary to account for the effect of our lower spectral resolution, which reach their maximum strength in red supergiants. The features which degrades the sharp drop in flux at the CO bandhead and of NaI (2.207 µm) and CaI (2.263 µm), also prominent in red super- spreads it toward shorter wavelengths in our spectra. The band- giants, are clearly visible in all the spectra. A technical problem affected pass we adopted thus ensures that the first CO band is fully en- the response of the detector during the observations of star 5, there- closed in it. fore the shape of the continuum of this star is not reliably represented The spectral types we estimate from EW(CO) are listed in in the spectrum. Table2. As shown in Fig. 2 of Davies et al. (2007) there is considerable spread in the EW(CO) vs. spectral type relation, Table 2. Equivalent widths of spectral features. especially among the latest-type supergiants. Spectral types estimated in this way can differ from the spectral classification in the visible by up to four or five subclasses. However, a comparison Star ID EW(NaI) EW(CaI) EW(CO) Sp. type (IR)1 between the spectral types estimated from EW(CO) and those ÅÅÅ listed in Table1 for the four stars for which classifications based 1 4.3 2.9 25.2 M1I on the visible spectrum exist in the literature indicate a fairly 2 4.8 3.5 23.7 M0I 3 4.5 3.5 31.1 M5I close match, with the largest discrepancy, in the case of Star 5, 4 4.8 4.1 23.9 M0I amounting to only two subclasses. 5 4.3 4.7 28.1 M3I 6 4.7 4.6 22.6 K7I 4.2. Space distribution 7 4.1 2.5 29.8 M4I Figure3 shows the location of the M supergiants projected on the (1) Notes. Spectral type estimated from the equivalent width of the first area of Cygnus OB2, and different classes of previously known CO(2–0) band. early-type stars indicated. The latter come from the survey of Comerón & Pasquali (2012), which is known to represent an in- complete census of early-type objects in the region. However, M supergiants can be distinguished from other cool lumi- the uniform photometric selection criterion used in that work nous stars through several spectroscopic features in the K-band leads to an approximately unbiased sample that is useful in the (Kleinmann & Hall 1986; Wallace & Hinkle 1996, 1997; Lançon present case for a comparison of the spatial distribution of cool & Wood 2000). The most outstanding one is the series of supergiants with those of other objects, particularly giant and su- CO bandheads starting at λ = 2.293 µm. This is characteris- pergiant B-type stars. Like other representatives of the evolved tic of cool photospheres, which reach their maximum depth in population of Cygnus OB2, we find all our cool supergiants to lie M supergiants. The features produced by some neutral species adjacent to the density peak dominated by O stars, in the region are also prominent, most notably NaI and CaI at 2.207 µm where evolved B giants and A stars (Drew et al. 2008) tend to be and 2.263 µm respectively. Table2 lists the equivalent widths located, thus confirming that star formation has been progressing

A46, page 4 of8 F. Comerón et al.: Red supergiants and the past of Cygnus OB2

Table 3. Derived extinction, temperature and luminosity.

Star ID AV Teﬀ (K) log L(L ) 1 14.4 3745 4.06 2 8.2 3790 4.02 3 10.9 3450 4.89 4 4.1 3790 3.95 5 5.8 3605 4.36 6 7.1 3830 4.17 7 11.0 3535 4.15

initial mass from Ekström et al.( 2012). For the effective temperatures we used the same calibration as in Davies et al. (2007), which is based on the spectral type vs. temperature calibration of Levesque et al. (2005) for Milky Way supergiants. The va- lidity of this calibration has recently been challenged by Davies et al. (2013), who have estimated effective temperatures of cool supergiants in the Large and Small Magellanic Clouds based on fits to MARCS model atmospheres (Gustafsson et al. 2008). Davies et al. (2013) find discrepant results when the temperature is derived based on either the fit to the TiO bands that dominate Fig. 3. Spatial distribution of the M supergiants discussed in this the spectroscopic classification in the visible, or to the overall study (filled circles). Other symbols represent stars from the survey of continuum. The authors also discuss shortcomings in the mod- Comerón & Pasquali (2012), which correspond to O-type stars (aster- eling of the TiO absorption that are likely to underestimate Teff isks), main sequence B stars (crosses) and B-type giants and supergiants when the latter are used for the fit. The problem is also indi- (open circles). The dashed line delimits the boundary of Cygnus OB2 as cated by Levesque et al. (2007), who found that the temperatures adopted by Comerón & Pasquali (2012) and is the area where we have searched for M supergiant candidates. For reference, the area studied in of Magellanic Cloud supergiants are significantly below evo- detail by Wright et al. (2015) is indicated by the smaller, dotted circle. lutionary model predictions at the appropriate metallicity. The comparison between Small and Large Magellanic Clouds presented by Davies et al. (2013) even suggests that the discrep- ancies increase with increasing metallicity, hinting at underes- toward higher Galactic longitudes in the time elapsed between timates of Teff that might reach over 600 K for early M-type the birth of the progenitor population of the currently observed supergiants in the Milky Way when temperatures are based on cool supergiants and the most recently formed early-type O stars. spectroscopic classifications in the visible. Since our spectral For reference, the area covered by the compilation of Wright type estimates are based on a calibration of the CO band depth et al. (2015) is also depicted in Fig.3. The latter includes a with respect to spectroscopic classifications in the visible as dis- smaller area where Wright et al. (2014) have studied the proper- cussed in Sect. 4.1, such caveats are equally applicable to the Teff ties of the lower-mass, X-ray emitting population, which is also vs. spectral type relation adopted for the spectral types listed in dominated by a very young (3–5 Myr) component. Table2. On the other hand, recent results on the determination of fundamental parameters of cool supergiants based on spatially 4.3. Evolutionary stage resolved interferometric observations tend to support the Teff vs. spectral type relation of Levesque et al. (2005) for nearby Both the age at which a star reaches the red supergiant phase, and Galactic red supergiants (e.g. Ohnaka et al. 2013; Arroyo-Torres the luminosity during that phase, are related to the initial mass of et al. 2013, 2015). By deriving Teff from the measured diameter its progenitor. Rotation and mixing in the interior of the star also at a variety of wavelengths, the dependency on the adopted at- have great importance in determining the post-main sequence mosphere model is greatly reduced with respect to the synthetic evolution and the path followed by the star in the temperature- atmosphere fits on which the method of Levesque et al. (2005) luminosity diagram (e.g. Ekström et al. 2012). is based. We adopted the average (J − K)0 color of M supergiants in Table3 lists the extinctions, adopted temperatures, and lumi- the compilation of Tokunaga (2000) to estimate the extinction nosities thus derived for our sample. We used AK = 0.0845AV in toward all our targets. Given the uncertainties in the classifica- agreement with Wright et al. (2015). We assumed an uncertainty tion of even previously known and spectroscopically classified temperature ∆Teff ∼ 200 K, which brackets the uncertainties in objects, together with the differences among different deriva- the Levesque et al. (2005) Teff vs. spectral type relation, as well tions of the average infrared colors for a given spectral type as an assumed uncertainty of two subclasses in the EW(CO) as discussed above, the resulting adopted value is (J − K)0 = vs. spectral type relation discussed in Sect. 4.1. The tempera- 0.93 ± 0.15, which translates into an uncertainty of 0.27 mag in ture uncertainty results in an uncertainty of '0.15 mag in BCK. the K-band extinction AK using the Cygnus OB2 extinction law However, given the importance of the latter quantity in deriv- from Wright et al. (2015). ing the initial mass of the red supergiant precursors, we adopted We derived K-band bolometric corrections BCK as a func- ∆BCK = 0.30 to account for systematic effects derived from a tion of the effective temperature Teff by using the values possible erroneous choice of the temperature scale. The adopted of the absolute bolometric magnitude Mbol, K-band absolute uncertainty in the luminosity is then computed as the quadratic magnitude MK, and effective temperature Teff from the solar- sum of the uncertainties in the photometry, the extinction, and metallicity evolutionary tracks of red supergiants with 10 M the bolometric correction.

A46, page 5 of8 A&A 586, A46 (2016)

20 5 15 6 7 1 12 2 10 4

Fig. 4. Location of our Cygnus OB2 M supergiant candidates with re- Fig. 5. Same as Fig.4, showing the e ﬀects of rotation for a M = 10 M spect to the Geneva evolutionary tracks, produced using the SYCLIST star in the proximities of the red supergiant phase. The curves corre- application (Georgy et al. 2014) and corresponding to an initial rota- spond to an initial rotational velocity v = 0.4vcrit (solid line), v = 0.9vcrit (dashed line), and a non-rotating model (dotted line). tional velocity v = 0.4vcrit (solid lines) and to non-rotating models (dotted lines). See Fig.5 for the identiﬁcation of the individual stars.

Figure4 shows the location of our seven stars in the of 20.9 Myr, and that rotation delays this phase until the star be- temperature-luminosity diagram, where Geneva evolutionary comes 25.2 Myr old if the star is rotating at 90% of vcrit. The tracks for models of different masses with and without rotation existence of such relatively low-luminosity red supergiants with (Ekström et al. 2012) are also plotted. For masses below ∼12 M initial masses of ∼10 M thus indicates that massive star forma- (the precise limit depends on the initial rotation rate), moderately tion was already proceeding in the region over 20 Myr ago. At rotating or non-rotating stars evolve to cool temperatures and higher masses, star 5 probably comes from a progenitor formed first become red supergiants at the beginning of the He burning between 15.4 Myr and 21.6 Myr ago, while that of star 3 formed phase, after which their temperatures temporarily increase, per- between 11.5 an 15.6 Myr ago. forming a blue loop in the temperature-luminosity diagram be- Our selection criteria should in principle be sensitive to red fore returning to low surface temperatures and ending their lives supergiants with Mbol > −5.15 and thus to masses below 10 M , as red supergiants again before they explode as type II super- but we did not find any candidates with these characteristics. novae. For a 10 M star, the duration of the red supergiant phase This might be an indication that the onset of star formation in previous to the blue loop, which we define as the time spent the region took place less than ∼25 Myr ago, thus implying that with a photospheric temperature Teff < 4000 K, lasts nearly stars with masses below 10 M , despite being more abundant 2 Myr (again depending on the initial rotation rate), which is by virtue of the shape of the mass function, have not had time much longer than the time spent in the final, post-blue loop red to reach the red supergiant phase. Given the small numbers in- supergiant stage. The evolution becomes simpler to describe for volved, the uncertainties in the derivation of the intrinsic proper- stars between ∼12 M and ∼20 M , which do not undergo the ties, and the uncertainties in the model evolutionary tracks this blue loop excursion. After reaching the red supergiant stage, conclusion cannot be firmly established, however, and needs to these stars remain there until the end of their lives, for a time be confirmed by future work. span ranging from somewhat less than 1 Myr to somewhat more than 2 Myr, depending on the initial mass and rotation velocity. An examination of the locations of our sample in the 4.4. Star formation rates present and past temperature-luminosity diagram suggests that the five least lu- A comparison between the catalog of early-type stars in the minous stars in our sample (numbers 1, 2, 4, 6, and 7) are lo- presently densest part of Cygnus OB2 produced by Wright et al. cated near the path followed by pre-blue loop cool supergiants (2015) and the number of cool supergiants that we find allows us with initial masses around ∼10 M , whereas star 5 probably has to obtain a rough estimate of the evolution of the star formation a more massive progenitor of ∼12 M and star 3 is the most mas- rate in the region. Wright et al. (2015) estimated their catalog to sive, with an initial mass of ∼15 M . The effects of different ini- be complete for stars with masses between 20 M and 40 M tial rotational velocities for initial masses of 10 M is illustrated and younger than 6 Myr. The individual masses and ages listed in Fig.5, where three tracks that correspond to no rotation, inter- by these authors show 31 stars with these characteristics in an mediate rotation, and rotation at 90% of the critical velocity vcrit area of one square degree considered in their study. Using the are illustrated. This shows that very fast rotation can suppress −2.3 initial mass function that they derive, ζ(M)M. ∝ M dM, this the blue loop as seen in Fig.5, but the duration of the cool su- implies the existence of ∼53 stars formed during the same inter- pergiant phase is still comparable to the pre-blue loop stage of val in the mass range 10 M < M < 15 M , where we find the slower rotating stars of the same initial mass. progenitor masses of the cool supergiants identified by us. This The presence of five relatively low-mass red supergiants has can be written as important consequences for the star formation history in or near Cygnus OB2, given the relatively long duration of their evo- N(10 M < M < 15 M , t < 6 Myr) = lution before they first enter the red supergiant domain. The Z 15 M Z 0 Geneva evolutionary tracks indicate that a non-rotating star with A1 ζ(M)dM Ψ(t)dt (1) initial mass of 10 M enters the red supergiant phase at an age M=10 M t=−6 Myr A46, page 6 of8 F. Comerón et al.: Red supergiants and the past of Cygnus OB2 where A1 is the area of 1 square degree considered by Wright 5. Conclusions et al. (2015), Ψ(t) is the average star formation rate per unit area, and we define t = 0 at present. The work presented here shows indications of an excess of very bright, red sources in the general direction of Cygnus OB2 that The number of stars with masses between 10 M and 15 M undergoing the red supergiant stage at present can be described we ascribe to the presence of red supergiants at the distance of through a similar expression: the association. Some of these sources have been found to be other types of objects by previous studies, but spectral classi- N (10 M < M < 15 M ) = fications in the literature complemented with new observations SG presented here show that seven such stars are indeed late-type Z v Z 15 M Z −t (M,v) crit 1 supergiants. The study of this sample leads us to the following A φ(v)dv ζ(M)dM Ψ(t)dt (2) 2 conclusions: v=0 M=10 M t=−t2(M,v) – The red supergiants lie in the region immediately adjacent where A2 is the area of our search (π square degrees); φ(v) is the distribution function of initial rotation velocities, which we to Cygnus OB2 and partly overlap it. In this region, where previous studies have revealed the presence of early-B stars assume to be uniform; and t1(M, v), t2(M, v) are the ages of a star of mass M and initial rotational velocity v at the time of that left the main sequence, as well as other types of stars entering and ending the red supergiant phase, respectively. The that indicate an average age older than that of the current evolutionary tracks at solar metallicity of Ekström et al. (2012) main concentration of O stars that form the youngest part of the association. The location of the reported population of were used to obtain t1(M, v) and t2(M, v). If we assume for simplicity that Ψ(t) has stayed constant at red supergiants is thus consistent with previous findings that have suggested that star formation has proceeded from lower Φ(t) = Φ1 over the past 6 Myr in which the young population characterized by Wright et al. (2015) formed, and that it had a Galactic longitudes toward the direction where the densest and youngest part of Cygnus OB2 is found at present. different value Φ(t) = Φ2 in the past when the progenitor popula- tions of the present red supergiants formed, the ratio of Eqs. (1) – The presence of red supergiants implies that star forma- and (2) yields tion in the Cygnus OB2 region started at least 20 Myr ago. Indications of a low-luminosity cutoff in this population of red supergiants suggest, however, that star formation started N(10 M < M < 15 M , t < 6 Myr) 53 Ψ1A1 = = 3.94 (3) not much earlier than this estimate. N (10 M < M < 15 M ) 7 Ψ A SG 2 2 – A comparison to the census of the younger (<6 Myr) population of Cygnus OB2 allowed us to carry out a crude esti- thus leading to Ψ ∼ 6Ψ , this is, the star formation rate in the 1 2 mate of the increase in the star formation rate in the region region has increased by a factor ∼6 since the progenitors of the with time. We find that the massive star formation rate has currently observed red supergiants were born. Furthermore, as increased by a factor of ∼6 between the present time (rep- noted in Sect. 4.2 the increasing star formation rate has been ac- resented by stars younger than 6 Myr) and the time when companied by a progression of the main star-forming locations the progenitors of the current population of red supergiants toward higher Galactic longitudes. The study of the spatial dis- formed. tribution and clustering properties of the youngest dynamically unevolved parts of the association carried out by Wright et al. Our results provide new support of the view that has emerged (2014) shows that recent star formation has resulted in relatively from studies carried out over the past decade. This view holds low-density aggregates, as opposed to clusters, with no indica- that the star formation history in the Cygnus OB2/Cygnus X re- tion of mass segregation. As noted by Wright et al. (2014), this gion started well before the last burst that gave rise about 3 Myr picture in which recent star formation has taken place at separate ago to the compact aggregate of early O-type stars that currently locations rather than as a single spatially confined burst suggests dominate the OB association. Our results extends this history that the star formation history in the Cygnus OB2 region has even further into the past. On the other hand, it must be kept been a succession of localized episodes extended over space, and in mind that the existence of neighboring OB associations such probably also in time although the study of Wright et al. (2014) as Cygnus OB1 and OB9, which lie approximately at the same is restricted by design to the youngest population. This picture is distance, have certainly also played a role in the overall evolu- consistent with the results we obtain when considering a much tion of the Cygnus region, thus highlighting the complexities of larger area, where the evidence for star formation extended in a star formation history that may be revealed by future studies of time becomes clear. tracers of its older stellar population. It may be noted that the duration of the star formation history in the Cygnus OB2 region that we infer is on the same order Acknowledgements. This paper is based on observations collected at the as the typical lifetime of giant molecular clouds in our Galaxy Observatorio del Teide, operated by the Instituto de Astrofísica de Canarias and others, which has been estimated to be around 20–30 Myr (IAC), in the island of Tenerife (Canary Islands, Spain); and at the Nordic Optical from numerical simulations and from direct observations (e.g. Telescope at the Observatory of Roque de Los Muchachos, in the Island of La Blitz et al. 2007; Murray 2011). On the other hand, the abundant Palma (Canary Islands, Spain). We are thankful to the astronomers in charge molecular gas reservoir still existing in the Cygnus X complex of service mode observing at the Observatorio del Teide for the careful execu- 6 tion of our program. F.C. warmly thanks the Specola Vaticana for their hospital- (∼10 M ; Schneider et al. 2006) that surrounds the Cygnus OB2 ity during the time when much of this work was done. F.C. and A.P. acknowl- association leads to the expectation that star-forming activity in edge support from the DFG Research Centre SFB 881 “The Milky Way System” it will still extend some Myr into the future. Taking this to- through project B5. The authors are grateful to the referee, Ben Davies, for very gether with the fact that massive star formation and its associated useful comments that helped improve the quality of this paper. This research has made use of the SIMBAD database operated at CDS, Strasbourg, France. It disruptive feedback were present already over 20 Myr ago, we also makes use of data products from the Two Micron All Sky Survey, which suggest that the parental giant molecular cloud whose remnants is a joint project of the University of Massachusetts and the Infrared Processing compose Cygnus X today may have had properties that led to a and Analysis Center/California Institute of Technology, funded by the National lifetime somewhat longer than typical. Aeronautics and Space Administration and the National Science Foundation.

A46, page 7 of8 A&A 586, A46 (2016)

References Knödlseder, J. 2003, in A Massive Star Odyssey: From Main Sequence to Supernova, eds. K. van der Hucht, A. Herrero, & C. Esteban (San Francisco: Arroyo-Torres, B., Wittkowski, M., Marcaide, J. M., & Hauschildt, P. H. 2013, ASP), IAU Symp., 212 A&A, 554, A76 Lançon, A., & Wood, P. R. 2000, A&AS, 146, 217 Arroyo-Torres, B., Wittkowski, M., Chaivassa, A., et al. 2015, A&A, 575, A50 Levesque, E. M., Massey, P., Olsen, K. A. G., et al. 2005, ApJ, 628, 973 Blitz, L., Fukui, Y., Kawamura, A., et al. 2007, in Protostars and Planets V., eds. Levesque, E. M., Massey, P., Olsen, K. A. G., et al. 2007, ApJ, 667, 202 B. Reipurth, D. Jewitt, & K. Keil, Univ. of Arizona Press Martins, F., & Plez, B. 2006, A&A, 457, 637 Brugel, E. W., & Wallerstein, G. 1979, ApJ, 229, L23 Massey, P., & Thompson, A. B. 1991, AJ, 101, 1408 Cameron, D., & Nassau, J. J. 1956, ApJ, 124, 346 Mengel S., & Tacconi-Garman, L. E. 2007, A&A, 466, 151 Clark, J. S., Negueruela, I., Davies, B., et al. 2009, A&A, 498, 109 Messineo, M., Qingfeng, Z., & Ivanov, V. D. 2014, A&A, 571, A43 Comerón, F., & Pasquali, A. 2012, A&A, 543, A101 Murray, N. 2011, ApJ, 729, 133 Comerón, F., Pasquali, A., Rodighiero, G., et al. 2002, A&A, 389, 874 Ohnaka, K., Hofmann, K.-H., & Schertl, D. 2013, A&A, 555, A24 Comerón, F., Torra, J., Chiappini, C., et al. 2004, A&A, 425, 489 Oliva, E., & Origlia, L. 1992, A&A, 254, 466 Comerón, F., & Pasquali, A., Figueras, F., & Torra, J. 2008, A&A, 486, 453 Rawlings, M. G., Adamson, A. J., & Whittet, D. C. B. 2000, ApJS, 131, 531 Davies, B., Figer, D. F., Kudritzki, R.-P., et al. 2007, ApJ, 671, 781 Reipurth, B., & Schneider, N. 2008, in Handbook of Star Formation Regions, ed. Davies, B., Kudritzki, R.-P., Plez, B., et al. 2013, ApJ, 767, 3 B. Reipurth, ASP Monograph (San Francisco: ASP) Drew, J. E., Greimel, R., Irwin, M. J., & Sale, S. E. 2008, MNRAS, 387, 08 Rygl, K. L. J., Brunthaler, A., Sanna, A., et al. 2012, A&A, 539, A79 Ducati, J. R., Bevilacqua, C. M., Rembold, S. B., & Ribeiro, D. 2001, AJ, 558, Schneider, N., Bontemps, S., Simon, R., et al. 2006, A&A, 458, 855 309 Straižys, V. 1987, Bull. Obs. Vilnius, 78, 43 Ekström, S., Georgy, C., Eggenberger, P., et al. 2012, A&A, 537, A146 Tokunaga, A. T. 2000, in Allen’s Astrophysical Quantities, ed A. N. Cox Georgy, C., Granada, A., Ekström, S., et al. 2014, A&A, 566, A21 (New York: AIP Press) Grasdalen, G. L., & Sneden, C. 1979, PASP, 91, 337 Uyaniker, B., Fürst, E., Reich, W., Aschenbach, B., & Wielebinski, R. 2001, Gustafsson, B., Edvardsson, B., Eriksson, K., et al. 2008, A&A, 486, 951 A&A, 371, 675 Haggqvist, L., & Oja, T. 1970, A&AS, 1, 199 Verhoelst, T., van der Zypen, N., Hony, S., et al. 2009, A&A, 498, 127 Hanson, M. M. 2003, ApJ, 597, 957 Wallace, L., & Hinkle, K. 1996, ApJS, 107, 312 Hirschi, R. 2010, ASP Conf. Ser., 425, 13 Wallace, L., & Hinkle, K. 1997, ApJS, 111, 445 Humphreys, R. M. 1978, ApJS, 38, 309 Wright, N. J., Drake, J. J., Drew, J. E., & Vink, J. S. 2010, ApJ, 713, 871 Imanishi, M., Sasaki, Y., & Goto, M. 1996, AJ, 112, 235 Wright, N. J., Parker, R. J., Goodwin, S. P., & Drake, J. J. 2014, MNRAS, 438, Kleinmann, S. G., & Hall, D. N. B. 1986, ApJS, 62, 501 639 Knödlseder, J. 2000, A&A, 360, 539 Wright, N. J., Drew, J. E., & Mohr-Smith, M. 2015, MNRAS, 449, 741

A46, page 8 of8